Exploring the Effects of ULEZ 2023 Expansion

Youran Xu

Gathering Input Data

Openair R package ¹
- Air Quality in England
- Automatic Urban and Rural Network
Supplementary Data from London-wide ULEZ Six Month Report 2024

Sites used in ULEZ report

176 sites for NO2
69 sites for PM2.5

198 sites from ULEZ report due to overlaps

## List of NO2 Monitoring Sites ####
sites_NO2 <- read_excel("Supplementary Data Sheet.xlsx", 
                        sheet = "List of NO2 Monitoring Sites")
sites_NO2 <- as.data.frame(sites_NO2)
## List of PM2.5 Monitoring Sites ####
sites_PM2.5 <- read_excel("Supplementary Data Sheet.xlsx", 
                        sheet = "List of PM2.5 Monitoring Sites")
sites_PM2.5 <- as.data.frame(sites_PM2.5)
## merge ####
sites_ulez <- full_join(sites_NO2, sites_PM2.5) ## remove overlaps between pollutants
sites_ulez <- sites_ulez %>% distinct() ## double check no duplicates
nrow(sites_ulez)

[1] 198

rm(sites_NO2, sites_PM2.5)
sites_ulez <- sites_ulez %>% 
  mutate(Name = ifelse(Name == "Beech Street NOX", "Beech Street NOx", Name))

Cross-referencing ULEZ sites and openair emission data

## AURN 
aurn <- importAURN(
  site = sites_ulez$`Site Code`,
  year = 2014 : 2024,
  data_type = "monthly",
  pollutant = c("nox", "no2", "no", "pm2.5", "pm10")
)
## AQE
aqe <- importAQE(
  site = sites_ulez$`Site Code`,
  year = 2014:2024,
  data_type = "monthly",
  pollutant = c("nox", "no2", "no", "pm2.5", "pm10")
)
## join
data <- aurn %>% 
  select(-3) %>% 
  full_join(aqe)
rm(aqe, aurn)
# 95 sites availible
data %>% 
  group_by(code) %>% 
  tally() %>% 
  nrow()

[1] 95

A total of 95 unique sites will be used in this analysis.

Transforming data set

## prepare variables for pivot
data <- data %>% 
  rename(no_concentration = no,
         no2_concentration = no2,
         nox_concentration = nox,
         pm2.5_concentration = pm2.5,
         pm10_concentration = pm10)
  
## pivot into long format
data <- data %>% 
  pivot_longer(
    cols = no_concentration:pm2.5_capture,
    names_to = c("pollutant", ".value"),
    names_sep = "_",
    values_drop_na = TRUE
  )

## only leave observations with a capture rate higher than 0.8
data <- data %>% 
  mutate(pollutant = factor(pollutant)) %>% 
  filter(capture >=0.8)
## add information about site type and area from ULEZ data set
data <- data %>% 
  left_join(sites_ulez, c("code" = "Site Code", "site" = "Name"))
# check for NAs
data %>%
filter(is.na(Area) | is.na(`Site Type`))

## add geographic coordinates from meta data
data <- data %>% 
  left_join(sites, by = join_by(code))
# check for NAs
data %>%
  filter(is.na(latitude))


## factorise variables
data <- data %>% 
  mutate(`Site Type` = factor(`Site Type`)) %>% 
  mutate(Area = factor(Area))

# adjust the order of area
table(data$Area)
data$Area <- factor(data$Area, 
                    levels = c("Central London",
                               "Inner London",
                               "Outer London",
                               "5 km from GLA Boundary",
                               "5-40 km from GLA Boundary"))
levels(data$Area)
data <- data %>% 
  mutate(source = factor(source))

saveRDS(data, file = "data.rds")

Data Overview

2 sources (AQE and AURN)
from Jan 2014 to October 2024
95 monitoring sites
Data capture rate above 0.8
pollutants: NO, NO2, NOX, PM10, PM2.5
18828 Roadside and 13118 Urban Background sites

::: aside Cited from Inner London Ultra Low Emission Zone – One Year Report 2023

:::

Observations of monthly average concentration for each pollutant:

Pollutant	NO. Observations
NO	7702
NO2	7698
NOX	7698
PM10	5516
PM2.5	3332

Calculating monthly % change for NOX

Code
Results preview
Average Percentage Change

mean_mon_pre <- data %>% 
  filter(pollutant == "nox") %>% 
  filter(date >= as.Date("2022-07-01") & date <= as.Date("2023-09-01")) %>% 
  group_by(date) %>% 
  summarise(mean_monthly = mean(concentration)) %>% 
  mutate(pct_change = (mean_monthly - lag(mean_monthly))/lag(mean_monthly) * 100)
  
mean_mon_pre %>% 
  filter(!is.na(pct_change)) %>%
  summarise(mean = mean(pct_change))

mean_mon_aft <- data %>% 
  filter(pollutant == "nox") %>% 
  filter(date >= as.Date("2023-09-01")) %>% 
  group_by(date) %>% 
  summarise(mean_monthly = mean(concentration)) %>% 
  mutate(pct_change = (mean_monthly - lag(mean_monthly))/lag(mean_monthly) * 100) 

mean_mon_aft %>% 
  filter(!is.na(pct_change)) %>%
  summarise(mean = mean(pct_change))

pct_change <- full_join(mean_mon_pre, mean_mon_aft)

Date	Average Monthly Concentration	Monthly % Change
Jul 22	34.94	/
Aug 22	37.52	7.37%
…	…	…
Aug 24	23.97	-17.34%
Sept 24	28.35	18.26%
Oct 24	31.01	9.40%

Before ULEZ 2023	After ULEZ 2023
4.16%	-1.15%

Monthly percentage change for NOX before & after ULEZ expansion 2023

Results

Graph
Timeline
Code

# data <- readRDS("data.rds")
theme_set(theme_minimal())
# overview ####
## all pollutants 2014-2024 ####
data %>% 
   ggplot(aes(x = as.Date(date), y = concentration, colour = pollutant)) +
   geom_smooth() +
   scale_x_date(date_breaks = "1 year", labels = date_format("%Y")) +
   theme(axis.text.x = element_text(angle = 30, hjust = 1)) +
   xlab("") +
   ylab("Monthly Average Concentration [μgm-3]") +
   ggtitle("Trends in monthly average air pollution concentrations in London", "From January 2014 to October 2024") +
   annotate("rect", xmin = as.Date("2023-09-01"), 
            xmax = as.Date("2024-10-01"), 
            ymin = -1, ymax = 150,
           alpha = .1,fill = "green") +
   annotate("text", x = as.Date("2024-04-01"), y = 130, label = "ULEZ", alpha = 0.8) +
   annotate("text", x = as.Date("2024-04-01"), y = 123, label = "Expansion", alpha = 0.8)

Let us zoom in and have a look:

Graph
Code

## all pollutants 2022-2024 ####

data %>% 
   filter(date >= as.Date("2022-10-01")) %>% 
   ggplot(aes(x = as.Date(date), y = concentration, colour = pollutant)) +
   geom_smooth() +
   scale_x_date(date_breaks = "3 month", labels = date_format("%Y %b")) +
   theme(axis.text.x = element_text(angle = 30, hjust = 1)) +
   xlab("") +
   ylab("Monthly Average Concentration [μgm-3]") +
   ggtitle("Trends in monthly average air pollution concentrations in London", "From October 2022 to October 2024") +
   annotate("rect", xmin = as.Date("2023-09-01"), 
            xmax = as.Date("2024-10-01"), 
            ymin = -1, ymax = 82,
           alpha = .1,fill = "green") +
   annotate("text", x = as.Date("2024-04-01"), y = 60, label = "ULEZ", alpha = 0.8) +
   annotate("text", x = as.Date("2024-04-01"), y = 55, label = "Expansion", alpha = 0.8)

Focus on NO2 after ULEZ expansion 2023…

Graph
Code

# By Area and site type ####
## NO2 ####
data %>% 
  filter(date >= as.Date("2022-07-01")) %>% 
  filter(pollutant == c("no2")) %>% 
  ggplot(aes(x = as.Date(date), y = concentration, linetype = `Site Type`)) +
  geom_smooth() +
  facet_wrap(~ Area) +
  scale_x_date(date_breaks = "8 months", labels = date_format("%y %b")) +
   theme(axis.text.x = element_text(angle = 30, hjust = 1)) +
   xlab("") +
   ylab("Monthly Average NO2 Concentration [μgm-3]") +
   ggtitle("Trends in monthly average NO2 concentrations in London", "From July 2022 to October 2024") +
  annotate("rect", xmin = as.Date("2023-09-01"), 
            xmax = as.Date("2024-10-01"), 
            ymin = -1, ymax = 50,
           alpha = .1,fill = "green") +
  geom_hline(aes(yintercept = 40, linetype = "dashed")) +
  scale_linetype(labels = c("Roadside", "Urban Background", "Limit Value"))

PM2.5 after ULEZ expansion 2023…

Graph
Code

data %>% 
  filter(date >= as.Date("2022-07-01")) %>% 
  filter(pollutant == c("pm2.5")) %>% 
  ggplot(aes(x = as.Date(date), y = concentration, linetype = `Site Type`)) +
  geom_smooth() +
  facet_wrap(~ Area) +
  scale_x_date(date_breaks = "8 months", labels = date_format("%y %b")) +
   theme(axis.text.x = element_text(angle = 30, hjust = 1)) +
   xlab("") +
   ylab("Monthly Average PM2.5 Concentration [μgm-3]") +
   ggtitle("Trends in monthly average PM2.5 concentrations in London", "From July 2022 to October 2024") +
  annotate("rect", xmin = as.Date("2023-09-01"), 
            xmax = as.Date("2024-10-01"), 
            ymin = -1, ymax = 20,
           alpha = .1,fill = "green") +
  geom_hline(aes(yintercept = 12, linetype = "dashed")) +
  scale_linetype(labels = c("Roadside", "Urban Background", "Environmental Improvement Plan"))

Conclusion

ULEZ 2023 expansion has been effective for its purpose.
Most effective for NO2 reduction in central London, outer London, and areas 5km from GLA boundary
Effective for PM2.5 reduction in central London
Effective for both roadside and urban background air quality
Results correspond with ULEZ six-month report from July.

Recommendations

The implementation of ULEZ zones could be promoted and generalised to other metropolitans.
Similar approach could be adopted for electric cars in the future.
Additional policies to improve urban air quality.
Meteorological data could be used to aid analysis in future reports.

Discussion

Technical:
- Limited sample size (95/198)
- Site name variations
- openair package

Methodological:
- Cross-comparison and chronological comparison
- Use of control group (estimation) and treatment groups could improve validity, as % change shows inconsistant results.
- Data smoothing reduces outliers from impacts of weather and seasonality, while losing data integrity and causing distortions.
Fit for trend-spotting but not impact analysis